| Title |
Design of a Physics-Informed Neural Network Based Thermal Model for Real-Time Core Temperature Prediction in Motor Dynamometer Systems |
| Authors |
정가은(Ga-Eun Jung) ; 이준엽(Jun-Yeop Lee) ; 이석주(Seok-Ju Lee) |
| DOI |
https://doi.org/10.5370/KIEE.2026.75.5.1124 |
| Keywords |
Core Temperature Prediction; Digital Twin; Motor Dynamometer System; Physics-Informed Neural Network (PINN); Reduced Order Thermal Model |
| Abstract |
This study proposes the design of a physics-informed artificial intelligence (AI) reduced-order thermal model for real-time core temperature prediction in motor dynamometer systems. The stator and rotor core temperatures of electric motors are difficult to measure directly and have traditionally been estimated using high-fidelity finite element method (FEM) simulations. However, the substantial computational burden of FEM-based approaches limits their applicability in real-time monitoring and control environments. To overcome this limitation, a lumped parameter thermal model that captures inter component heat transfer, ambient temperature effects, and current dependent loss mechanisms is integrated with a physics-informed neural network (PINN) framework. The proposed method embeds the governing thermal dynamics into the learning process, enabling the simultaneous identification of neural network weights and physically interpretable thermal parameters through a composite loss function that minimizes both temperature data mismatch and physics residuals. Furthermore, the model is structured in an ordinary differential equation (ODE) form to ensure numerical efficiency and stability under varying operating conditions. Extensive validation under multiple load profiles demonstrates that the proposed approach preserves FEM level prediction accuracy while achieving about five orders of magnitude reduction in computation time. Consequently, the developed model provides a practical foundation for real-time thermal monitoring, digital twin implementation, and advanced performance evaluation of motor dynamometer systems in industrial applications. |